Journal of Medicinal C h i s t r y

0 Copyright 1977 b y the American Chemical Society

Volume 20, Number 4 April 1977

A Manual Method for Applying the Hansch Approach to Drug Design’

John G. Topliss
Department of Chemical Research, Schering-Plough Research Division, Bloomfield, New Jersey 07003.
8eceiued August 18, 1976

A procedure is described in which an initial small group of compounds is selected, tested, and ordered according
to potency. The potency order in the group is then compared to the tabulated potency order calculated for various
parameter dependencies relating to hydrophobic, electronic, and steric effects. From this activity pattern analysis
the probable operative parameters can be deduced and a new substituent selection made for the Synthesis of potentially
more potent analogues. Application of the method is illustrated with a series of examples. It differs from a previously
described decision tree, single compound stepwise approach in that it involves the batchwise analysis of small groups
of compounds, usually the preferred procedure for logistical reasons if the compounds are relatively easy to synthesize.

A very common problem in drug design is to find the
optimum substitution on a benzene ring or on the ben-
zenoid portion of a fused ring system in an active lead
compound for maximization of drug potency. A literature
survey has shown that some 40% of all reported com-
pounds incorporate an unfused benzene ring.3 Also more
than 50% of drug-oriented patents are concerned with
substituted benzene^.^ Efficient solutions to the problem
of optimizing substitution would therefore have much
value. With the advent5 and subsequent development6 of
the Hansch method for structure-activity correlations, a
more rational approach to this problem became possible.
Subsequently, a number of publications7-” have dealt with
the most advantageous procedural strategies for sub-
stituent selection in applying the Hansch method.
In a previous publication the utilization of operational
schemes for analogue synthesis in drug design was de-
scribed.2 This method, which is based on the fundamental
assumptions of the Hansch approach, involves the stepwise
selection for synthesis of new analogues of an active lead
compound designed to maximize the chances of syn-
thesizing the most potent compounds in the series as early
as possible. The stepwise selection takes the form of a
decision tree and does not require multiple regression
analysis. Application of the method should be particularly
advantageous when analogue synthesis is difficult and slow
and test results are relatively rapid. It is also of interest
as a way of applying Hansch type principles without the
use of statistical procedures and computers.
The present paper describes a procedure, not involving
statistical methodology, by which groups of analogues
representing different types of substitution on a benzene
ring can be selected and synthesized at one time. As such
it may be viewed as a manual method for applying the
Hansch approach to drug design.
The initial group of analogues selected for synthesis
consists of the first five compounds in Table I. This set
of five compounds comprises the top section of the op-
erational scheme for aromatic substitution previously
463
described2and the compounds should be readily accessible
from a synthetic standpoint.
The projected order of potency of these five compounds
for various parameter dependencies is listed in Table 11.
Comparison with the actual experimentally determined
potency order allows a possible deduction to be made
concerning the probable operative parameters which in
turn provides the basis for a new substituent selection from
Table 111. The new substituent selections are examples
of suitable choices and clearly other selections can be made.
An examination of Tables I and I1 reveals that the ex-
pected potency order for some parameter dependencies is
quite similar. Given the fact that only five compounds are
being considered and taking into account some possible
uncertainty in the biological data it is apparent that a
narrowing of the probable operative parameters to a
related group of possibilities may be accomplished rather
than a precise identification. These related groups of
probable operative parameters are listed in the left-hand
column of Table 111. It should also be clear that the bi-
ological potencies of the initial compound group have to
show sufficient spread to permit a meaningful analysis.
The parameter dependencies listed in Table I1 en-
compass those most commonly found in Hansch type
correlations, viz.linear and parabolic P,a, and a wide range
of P f a. Steric effects, E , are also included where these
exert a dominant influence. The choice of coefficients in
the various P f a dependencies reflects a range of relative
weightings of the two parameters. Different relative
weightings are included only if they produce a significant
change in the potency order for the five compounds in
Table 11. Thus, the potency order for P, a, and P + a is
very similar and there would be no point in including
columns for 2?r + a, P + 2a, etc. On the other hand, the
potency order for P and -a is essentially reversed, hence
the opportunity to consider not only P - a but 27r - a, P
- 2a, and P - 30.
At this stage it is appropriate to illustrate the application
of the method with some examples. In reviewing the
464 Journal of Medicinal Chemistry, 1977, Vol. 20, No. 4 Topliss
Table I. Parameter Valuesa
Substituents n 5 n-u 2n-u n-20 n-35 -5 7' 7 f U 2n-n2
H 0 0 0 0 0 0 0 0 0
4-C1 0.71 0.23 0.48 1.19 0.25 0.02 - 0.23 0.94 0.92
3,441, 1.25 0.52 0.73 1.98 0.21 -0.31 - 0.52 1.77 0.94
4-CH3 0.56 - 0.17 0.73 1.29 0.90 1.07 0.17 0.39 0.81
4-OCH3 - 0.02 -0.27 0.25 0.23 0.52 0.79 0.27 - 0.29 - 0.04
3-CF3,4-C1 1.59 0.66 0.93 2.52 2.91 - 0.39 ~-0.66 2.25 0.65
3-CF3,4-NO, 0.60 1.21 -0.61 ~~0.01 - 1.82 - 2.97 - 1.21 1.81 0.84
4-CF3 0.88 0.54 0.34 1.22 - 0.20 - 0.74 - 0.54 1.42 0.99
2,4 -C1, 1.42 0.46 0.96 2.38 0.50 0.04 - 0.46 1.88 0.82
C-C H, 2.14 - 0.02 2.16 4.30 2.18 2.20 0.02 2.12 - 0.30
c-CJ-4, 2.51 - 0.22 2.73 5.24 2.95 3.17 0.22 2.29 - 1.28
4-CH(CH3), 1.53 - 0.05 1.58 3.11 1.63 1.68 0.05 1.48 0.72
4-C(CH3), 1.98 - 0.20 2.18 4.16 2.38 2.58 0.20 1.78 0.04
3,4-(CH,), 0.99 -0.30 1.29 2.28 1.59 1.89 0.30 0.69 1.00
4-O(CHZ),CH, 1.55b -0.32 1.87 3.42 2.19 2.51 0.32 1.23 0.70
4-OCH2Ph 2.13b -0.42 2.55 4.68 2.97 3.39 0.42 1.71 - 0.28
4-N(C2H,), 1.18 - 0.83 2.01 3.19 3.84 4.67 0.83 0.35 0.97
I-N(CH,)z 0.18 .- 0.83 1.01 1.19 1.84 2.67 0.83 - 0.65 0.33
4-NH2 4.23 - 0.66 -0.57 ~ 1.80 0.09 0.75 0.66 -1.89 - 3.97
4-NHC,H9 1.45 -0.51 1.96 2.39 2.47 2.98 0.51 0.94 0.80
4-OCH(CH,), 1.03b - 0.45 1.48 2.51 1.93 2.38 0.45 0.58 1.00
3-CH3,4-OCH, 0.54 -0.26 0.80 1.34 1.06 1.32 0.26 0.28 0.79
4-Br 0.86 0,23 0.57 1.49 0.40 1.55 -0.23 1.09 0.98
3-CF3 0.88 0.43 0.45 1.43 0.02 2.17 -0.43 1.31 0.99
4-C,H, 1.02 -0.15 1.17 2.19 1.32 1.47 0.15 0.87 1.00
4-O(CH2),CH, 1.05 -0.25 1.30 2.35 1.55 1.80 0.25 0.80 1.00
3-CH3,4-C1 1.29 0.17 1.12 2.41 0.95 0.78 -0.17 1.46 0.92
3-C1 0.71 0.37 0.34 1.05 -0.03 - 0.30 -0.37 1.08 0.92
3-CH3 0.56 - 0.07 0.63 1.19 0.70 0.77 0.07 0.49 0.81
3-OCH3 -0.02 0.12 - 0.14 -0.16 - 0.26 - 0.38 -0.12 0.10 --0.04
3-N(CH3), 0.18 -0.15 0.33 0.5 1 0.48 0.63 0.15 0.03 0.33
3,5-C1, 1.25 0.75 0.50 1.75 -0.25 - 1.00 -0.75 2.20 0.94
2-ci 0.71 0.23 0.48 1.19 0.25 0.02 -0.23 0.94 0.92
2-CH3 0.56 -0.17 0.73 1.29 0.90 1.07 0.17 0.39 0.81
2-OCH3 -0.02 - 0.27 0.25 0.23 0.52 0.79 0.27 -0.29 - 0.04
2-F 0.14 0.06 0.08 0.22 0.02 - 0.04 -0.06 0.20 0.28
4-F 0.14 0.06 0.08 0.22 0.02 -0.04 -0.06 0.20 0.28
4-NHC OCH - 0.97 0.00 -0.97 - 1.94 -0.97 -0.97 0.00 -0.97 - 2.88
4-NHSOzCH, -1.18 0.03 --1.21 -2.39 - 1.24 - 1.27 - 0.03 - 1.15 - 3.75
4-NO, -0.28 0.78 - 1.06 - 1.34 - 1.84 - 2.62 - 0.78 0.50 - 0.64
4-COCH3 -0.55 0.50 - 1.05 - 1.60 - 1.55 - 2.05 -0.50 -0.05 - 1.40
4-S02CH3 - 1.63 0.72 - 2.35 -3.98 - 3.07 - 3.79 -0.72 -0.72 - 5.92
4-CONHz - 1.49 0.36 - 1.85 -3.34 - 2.21 - 2.57 -0.36 -1.13 - 5.20
4-SOzNH, -1.82 0.57 - 2.39 -4.21 -3.01 - 3.53 - 0.57 -1.25 -6.95
a C. Hansch, A. Leo, S. H. Unger, K. H. Kim, D. Nikaitani, and E. J. Lien, J . Med. C h e m . , 16, 1207 (1973). Estimated
value.

Table 11. Potency Order for Various Parameter Dependencies

Parameters
Substituents rr 2H -- R2 n t 2n-0 a-25 n- 30 E4a
__ CI -5 R - U

3,4-C12 1 1-2 1 5 1 1 1-2 3-4 5 2-5

4-C1 2 i-2 2 4 2 2-3 3 3-4 3-4 2-5
4-CH3 3 3 4 2 3 2-3 1-2 1 1 2-5
4-OCH, 4- 5 4-5 5 1 5 4 4 2 2 2-5
H 4-5 4 4 3 3 4 5 5 5 3-4 1
_____I___

Unfavorable steric effect from 4 substitution.

Table 111. New Substituent Selection

Probable operative
parameters New substituent selection
_. __ _______
n, n t 0 , 0 3-CF3, 4-Cl; 3-CF,, 4-NO,; 4-CF,; 2,4-C1,; 4-c-C5H,;4-c-C6H,,
", 211 ~ 5 , K ~- 0 4-CH(CH3),;4-C(CH3),;3,4-(CH,),; 4-O(CH,),CH3;4-OCH2Ph;4-N(C,H5),
TT -- 20, K -. 30, --a 4-N(C,H5),;4-N(CH3),;4-NH2;4-NHC,H9;4-OH; 4-OCH(CH3),;3-CH,,4-OCH3
2n ~~ 712 4-Br; 3-CF3;3,4-(CH,),; 4-C,H5;4-O(CH,),CH3;3-CH3,4-C1
3-Cl; 3-CH3; 3-OCH3;3-N(CH3),;3-CF,; 3,5-C1,
Ortho effect 2C1; 2-CH3;2-OCH3; 2-F
Other 4-F; 4-NHCOCH,; 4-NHSO,CH,; 4-NO. ; 4-COCH,; 4-SO,CH,; 4-CONH,; 4-SO,NH,

results of these examples criteria for judging the degree parameter dependency as determined by a standard
of success of the method are (a) rapid identification of the Hansch analysis on a larger series. The first example
more potent analogues and (b) deduction of the correct concerns the antiinflammatory activity of some 5-aryl-
Application of Hansch Approach to Drug Design Journal of Medicinal Chemistry, 1977, Vol. 20, No. 4 465
Table IV. Antiinflammatory Activity of Table VI. Sulfonamide Carbonic Anhydrase Inhibitorsa
5-AryltetrazolylpropionicAcidsa

Rank order
Rank order Calcd
Log
X AIb Obsd Calcd for E , X 1IK.b Obsd n n +B
Initial Compound Group Initial Compound Group
3,4-C1, 6.2 2 2-5 3 ,4-C1, 1.40 1 1 1
4-C1 5.9 3 2-5 4-C1 0.72 2 2 2
4-CH, 3.1 5 2-5 4-CH3 0.42 3 3 3
4-OCH3 4.9 4 2-5 4-OCH3 0.35 4 4-5 5
H 8.2 1 1 H 0.22 5 4-5 4
Second Compound Group Second Compound Group
3-C1 11.2 1-3 3-CF3,4-N0, 1.85
3-CH3 7.9 4 Reference 14. K , represents carbonic anhydrase in-
3-CF3 5.7 5 hibition constant. The value 1 / K , is proportional to the
3,5-C1, 11.2 1-3 magnitude of carbonic anhydrase inhibitory activity.
Other Compounds
3-Br 11.2 1-3 was the most potent compound in the 22 compound series.
3-NH, 0.3 6 By means of a conventional Hansch treatment using
a Reference 12. Activity index = ten times the ratio multiple regression analysis, Fuller et al.I3 established that
of the mean pleural exudate volumes of the test com- activity correlated in a positive sense with H + u. Thus,
pound at 1 mmol/kg over 1 mmol/kg of aspirin. the conclusion with respect to parameter dependency
reached by analysis of the initial five compound group was
Table V. Inhibition of PNMT by 3- and correct. It is important to recognize that it is not usually
4-Substituted Amphetaminesa necessary to know the precise parameter dependency in
order to select analogues which have a good probability
of showing increased potency. Thus, the analogue se-
quence H; 4-C1; 3,4-Cl2; 3-CF3,4-C1; 3,4-(CF3)2,having
progressively higher values for both ZH and Xu, would
Rank order produce progressively more potent analogues for H, u, or
Calcd for H + u dependencies.
X Obsd n +u Kakeya et al.14 reported structure-activity studies on a
Initial Compound Group series of 19 substituted sulfonamides which show carbonic
3,4421, 5.10 1 1 anhydrase inhibitory activity. Tabulation of the initial
4-C1 3.60 2 2 compound group (Table VI) and rank ordering according
4-CH, 3.14 3 3 to biological potency indicated a ?r or H + u relationship
4-OCH, 2.57 5 5 (Table 11). This is consistent with Kakeya et al. analysis
H 2.89 4 4 of the entire 19 compound series using regression analysis
Second Compound Group which showed that activity was H + u (primarily a) related.
4-CF3 4.00 Of the indicated second compound group only one
a Reference 13. Negative log of the molar concen- member, the 3-CF3,4-N02analogue, had been synthesized.
tration of inhibitor needed for 50% inhibition of enzyme However, this compound was the most potent in the entire
activity. series.
It is not actually necessary to have information on all
tetrazolylpropionic acids12(Table IV). In this series rank five compounds in the specified initial compound group.
ordering according to potency of the initial compound Often useful conclusions can be drawn on only four
group reveals the parent phenyl compound to be sub- compounds. For example, in a series of 13 N‘-benzoyl-
stantially more potent than the others. Reference to Table ~ulfanilamidesl~ with antibacterial activity against Es-
I1 indicates the inference to be drawn is that there is cherichia coli, the 3,4-C12analogue is not reported. Or-
probably an adverse steric effect at position 4 and that a dering the four available compounds in the initial com-
second group of compounds should be synthesized and pound group as shown in Table VII leads to the conclusion,
tested with substituents at the 3 and 3,5 positions. Among by reference to Table 11, that activity is probably -u
the indicated second compound group shown in Table IV dependent. This is the same as the conclusion reached by
are three of the four most potent analogues in the 28 Cammaratal’ based on a regression analysis of the 13
compound series. compound series. In the indicated second compound group
The method may be further illustrated by reference to taken from Table 111, the 3-CH3,4-OCH3analogue had
a series of 3- and 4-substituted amphetamines which are been made and was in the high potency category. Other
inhibitors of phenethanolamine N-methyltransferase compounds, which formed part of the reported series and
(Table V). Rank ordering by potency of the initial expected to show high potency based on a -u dependency,
compound group suggests a H + u or possibly a ?r or u are the 3-CH3,4-C2H5,4-n-C3H7,and 4-i-C3H7analogues.
dependency (Table 11). The next step would be the The observed potencies of these are consistent with this
synthesis and testing of the second compound group (Table projection.
VI. Only one compound from this group was reported by A n interesting case is the structure-activity relationships
Fuller et al.13 but its potency was in line with the parameter of g-(X-phenyl)guanine inhibitors of xanthine oxidase
dependency indicated by analysis of the first compound reported by Silipo and Hansch.16 Analysis of the initial
group. The 3,4-C12analogue included in the first group five compound group taken from this series of 33 com-
466 Journal of Medicinal Chemistry, 1977, Vol. 20, No. 4 Topliss
Table VII. In Vitro Activities of Benzoylsulfanilamides Table IX. Inhibition of Hill Reaction by
against Escherichia coli a Phenyldimethylureasa

Rank order Rank order

Logb Calcd Calcd
X 1IC Obsd for-a X Log l / C b Obsd forn + a
Initial Compound Group Initial Compound Group
441 5.10 4 4 3,4-C1, 6.70 1 1
4-CH3 5.40 1-2 2 4-C1 5.40 2 2
4-OCH, 5.40 1-2 1 4CH, 4.52 3 3
H 5.25 3 3 4-OCH3 4.30 5 5
H 4.40 4 4
Second Compound Group
3-CH3,4-OCH, 5.25 Reference 17. C is the concentration which pro-
duces 50% inhibition of the Hill react.ion.
Other Compounds
3-CH, 5.40
4-C2H, 5.62 Table X. Inhibition of Guanine Deaminase by
4-n-C3H, 5.18 9-(X-Phenyl)guaninesa
4-i-C3H, 5.40 OH
3-CF3 4.65 I
4-NO, 4.50
4-CN 4.05
a Reference 15. C is the minimum inhibitory concen-
tration against E. coli.

Table VIII. 9-(X-Pheny1)guanine Inhibitors of

Xanthine Oxidasea Rank
X Log l / C b order obsd
Initial Compound Group
3,441, 6.03 2
4-C1 5.42 3-4
4-CH3 5.45 3-4
4-OCH, 6.70 1
H 5.00 5
Second Compound Group
Rank order 4-0C, H 7.01
4-O(CH2),C6H, 7.25
X Log l / C b Obsd Calcd for E ,
a Reference 18. C represents the molar concentration
Initial Compound Group for 50% reversible inhibition of guanine deaminase from
3,4-C1, 5.96 3 2-5 rabbit liver when assayed at pH 7.4.
4-C1 5.74 4-5 2-5
4-CH3 5.80 4-5 2-5
4- OCH 6.30 1-2 2-5 correlated with the combined molecular refractivity of
H 6.39 1-2 1 substituents at the 3 and 4 positions.
Second Compound Group A series of herbicidal phenyldimethylureas which inhibit
3-C1 6.57 photosynthesis and originally studied from a structure
3-CH3 6.62 activity standpoint by Hansch and Deutsch" contains all
3-OCH, 6.66 five compounds needed for the initial compound group
3-CF3 6.82 analysis (Table 1x1. The rank order of these analogues
a Reference 16. C is the concentration which pro- corresponds with the expected order for a u + u parameter
duces 50% inhibition of xanthine oxidase. dependency, consistent with the findings of Hansch and
Deutsch arrived at by multiple regression analysis on the
pounds (Table VIII) strongly suggests that there is an 12 member series. The indicated new substituent selection
adverse steric effect from 4 substitution. Thus, the two (Table 111)could not be checked since these compounds
most potent members of this group, the 4-unsubstituted were not reported in the original study.
and the 4-OCH3 compounds, have the lowest steric re- Even where the rank order of the substituents with
quirements, whereas the two least potent, the 4-CH3and respect to activity does not correspond completely to any
4-C1 analogues, have the highest steric requirements as parameter dependency listed in Table 11, it may be still
measured by E, values. The intermediate potency value possible to conduct a useful analysis leading to the
of the 3,4-C12 analogue is consistent with the idea of a identification of more potent analogues. This is illustrated
positive effect on activity of the 3-chloro substituent. Of by reference to a series of 9-(X-pheny1)guanines which
the second compound group specified in Table ID, the 3421, inhibit guanine deaminase (from rabbit liver)l8(Table X).
3-CF3,and 3-CH3analogues had been reported by Silipo From Table I1 no combination of parameters can account
and Hansch and they were all more potent than any of the for the 4-OCH3 analogue having the highest activity and
compounds in the initial compound group. From a 3,4Cl2the second highest. Therefore, it must be concluded
multiple regression analysis on the entire series, Hansch that special factors are operative which are not part of the
and Silipo concluded that activity was negatively correlated assumptions made in constructing Table 11. Leaving out
with substituent size at the 4 and 2 p i t i o n s and positively temporarily the 4-OCH3 compound from consideration, the
Application of Hansch Approach to Drug Design Journal of Medicinal Chemistry, 1977, Vol. 20, No. 4 467
Table XI. Hypoglycemic Activity of Table XII. Complex Formation between X-C,H,CONH,,
Phenacyltriphen ylphosphoranesa Alcohol Dehydrogenase, and DPNHa

Rank order Rank order

Hypoglycemic Calcd for Log Calcd
X act. Obsd E4 X KERJ Obsd for n - (I
Initial Compound Group Initial Compound Group
3,441 2 3-5 2-5 4-C1 - 1.93 2 2
4-C1 - 21 1-2 2-5 4-CH3 - 1.78 1 1
4-CH3 -2 3-5 2-5 4-OCH, - 2.20 3 3
4-OCH, -4 3-5 2-5 H - 2.72 4 4
H - 24 1-2 1
Second Compound Group
Second Compound Group 4-CH(CH3), -1.70
3-C1 - 42
3-CH3 - 30 a Reference 20. K E R I is the dissociation constant
3-OCH, - 42 for the complex of alcohol dehydrogenase with benz-
3-CF3 - 22 amides and DPNH.
a Reference 19. Expressed as the percent difference
between the mean change in control and treated groups Table XIII. Fibrinolytic Activity of
(rats) at 4 h. 2-Phene thynylcyclopropanecarboxylatesa

remaining four analogues in the group display a relative

potency pattern consistent with a + P dependency. One
consideration is that the 4-OCH3 analogue activity de-
termination might be in error. However, in order to fit the Rank order
+T dependency idea, the magnitude of the experimental X Log l/Cb Obsd Calcd for n
error would have to be very large (from 6.70 to ca. 5.0), ~~ ~~~ ~~

which seems improbable. A second consideration would Initial Compound Group

be that the 4-OCH3 group confers a large activity en-
hancement through a factor or factors which are not part
of the assumptions made in constructing Table 11. As-
suming this to be the case, indicated new analogues for
synthesis would be other 4-alkoxy and also 4-aralkoxy
compounds where this same special factor would probably
also be operative. Higher P values associated with the new
4-alkoxy compounds might also be expected to have a
positive effect on activity. Two compounds of this type,
the 4-OCzH5and 4-0(CHz)3C6H5,were reported in the
original study and had the third and second highest ac-
tivities, respectively, in the entire 33 compound series.
Other choices, not reported, would be the 4-0(CH2)&H3,
4-OCH(CH3)2, 4-0CH2C6H5,and 4-0(CH2)zC6H5ana-
logues. Silipo and Hanschls concluded, on the basis of a
detailed study of the series using multiple regression
analysis and surveying 1920 equations, that activity was
positively correlated with the P value of the 4-substituent,
the molar refractivity of the 3-substituent, and an indicator
variable associated with the presence of a 4-alkoxy sub-
stituent, and negatively correlated with the steric constant
of the 2-substituent. The present analysis, although not
providing this degree of understanding of structure-ac-
tivity relationships, nevertheless pointed the way to some
of the most potent compounds in the series in a relatively
simple manner while requiring consideration of only a very
limited number of compounds.
Another illustration of the use of the method can be
made from the publication of Blank et al.19 on the hy-
poglycemic activity of a series of phenacyltriphenyl-
phosphoranes. Although the rank order of the initial
compound group (Table XI) does not exactly correspond
to any parameter dependency given in Table 11, it is more
consistent with the idea that there is probably some
unfavorable steric effect from 4 substitution in view of the
rank of the 4-H compound. This, leads from Table I11 to
the selection of the second compound group shown in
Table XI. Of this latter group data on four compounds
4-C1 1.92 1 1
4-CH3 1.82 2 2
4-OCH3 1.70 3-4 3-4
H 1.70 3-4 3-4
a Reference 21. C is the molar concentration of anion
which dissolves a standard clot of human plasma in 24 h.
are available and show a definite trend to substantially
increased potency. The two most potent compounds in
the series are included in this group.
In a series of benzamides which inhibit alcohol de-
hydrogenase,z0the 3,4-Cl2analogue is missing from the
specified initial compound group. Rank ordering of the
four available compounds as shown in Table XI1 suggests
(Table 11) a A - u activity dependency for the series. Of
the indicated second compound group (Table 111) only the
4:CH(CH3)2 was reported and this was the most potent
member of the series. It is noteworthy that Hansch et aL20
established by regression analysis on the 15 compound
series that activity correlated with a A - u function.
A series of 2-phenethynylcyclopropanecarboxylates
exhibiting fibrinolytic activityz1 provides another illus-
tration of how, from rank ordering four compounds, it is
possible to make a useful judgement concerning probable
parameter dependency (Table XIII). By reference to
Table I1 a probable linear dependence of activity on T is
indicated. This turns out to be consistent with the
conclusion reached by Yoshimoto et aLZ1who found ac-
tivity to be a linear function of log P by a regression
analysis on the nine compound series.
A further example where four compounds of the initial
compound group are available is a series of 2-phenyl-
l,&indandiones which are uncouplers of oxidative phos-
phorylation in rat liver mitochondria.22 Rank ordering by
potency of the four compounds (Table XIV)and reference
to Tables I and I1 suggests that activity is A dependent
with the possibility of a small -Q effect. This leads (Table
111)to the choice of the second compound group which
468 Journal of Medicinal Chemistry, 1977, Vol. 20, No. 4 Topliss

Table XIV. Uncoupling Activities of Table XV. Antiallergic Activity of

2-Phenyl-1,3-indandionesa 2-Phenyl-8-azapurin-6-onesa
0 0

Rank order
Rank order
Calcd for
X Log 1 / C S o b Obsd 2n - a X Re1 act.b Obsd Calcd for E ,
Initial Compound Group Initial Compound Group
4-C1 4.20 1-2 1-2 4-C1 2 2 2-4
4-CH3 4.10 1-2 1-2 4-CH, 0.8 3-4 2 -4
4-OCH, 3.75 3-4 3 4-0CH3 1 3-4 2-4
H 3.74 3-4 4 H 4 1 1

Second Compound Group Second Compound Group

4-CH(CH3), 4.87
4-C(CH,), 5.03
a Reference 22. C,, = concentration for 50% uncou-
pling.
contains two compounds, the 4-CH(CH3)2and 4-C(CH3)3
analogues, where information is available. These both
showed greatly enhanced potency over compounds in the
initial compound group. By application of multiple re-
gression analysis to the entire 44 member series van den
Berg et a1.22concluded that x was the most important
parameter controlling activity with a -u effect playing a
secondary role. Ortho substitution was shown to be a
negative influence.
Situations will sometimes arise where after following the
procedure through the second compound group phase no
clear-cut progress toward identifying analogues with
improved activity will have been made. In such cases an
examination of 2-substituents is indicated as a next step.
An example is provided by a series of 2-phenyl-8-aza-
purin-6-ones showing antiallergic activity.23From a study
of the relative activities of the initial compound group
(Table XV), it is clear that 4 substitution is unfavorable.
This suggests an examination of 3-substituents (Table 11)
for the second compound group. Here no advantage is
gained in activity over the parent compound which leads
to the 2-substituted analogues in the third compound
group. Of this group the 2-OCH3 compound has excep-
tional activity compared to the 2-C1 and 2-CH3analogues
and is considerably more active than the parent compound.
These relative activities cannot be explained on the basis
of R , u, and E, influences and it is clear from this that some
special factor is involved in explaining the high potency
of the 2-OCH3analogue and it would be obvious from this
point to explore other 2-alkoxy compounds. The 2-alkoxy
compounds are, in fact, the most active group in the entire
44 compound series with the 2-n-propoxy compound
showing peak activity. Wooldridge and c o - ~ o r k e rwere s~~
able to show that the special factor involved was probably
a strong positive activity influence from the hydrogen
bonding capacity of the ortho substituents and by a
multiple regression technique were able to establish an
equation relating activity quantitatively to this positive
factor and negatively to the bulk of the 2-substituent. It
is interesting to note that the special potential of the
2-OCH3 is recognized after an examination of ten com-
pounds in the present analysis which can be considered
quite efficient considering the unusual aspects involved.
Wooldridge and c o - ~ o r k e r reached
s~~ the same stage in
their analysis after an attempted correlation using multiple
regression analysis with partition, electronic, and steric
H 4 1-3
3-CH3 4 1-3
3-OCH3 2 5
3-N(CH3)2 0.5 4
3-CF3 4 1-3
Third Compound Group
H 4 2
2-c1 0.2 4
2-CH3 0.04 5
2-OCH3 10 1
2-F 0.5 3
a Reference 23. * Activity relative to disodium cromo-
glycate (=1)in the rat PCA test following iv administra-
tion.
parameters on an initial ten compound series.
As illustrated in the last example, the examination of
2-substituents is made after first exploring the potential
of 4- and 3-substituents. For most series there is a priori
no reason to expect that one position of substitution will
be superior to another as far &s activity is concerned. Thus,
the choice of sequence can logically be made on the basis
of synthetic and economic considerations which generally
favor 4 and 3 over 2 substitution. In the exceptional cases
where there is a strong rationale for expecting favorable
effects on activity from a particular substituent position
the sequence should be modified accordingly.
In Table I11 the new substituent selection listed for the
designation “other” under probable operative parameters
requires comment. These are indicated for investigation
if the analysis, through the phase of examination of 2
substitution, has not yielded compounds of substantially
increased potency. The 4-F analogue, which provides
minimal change in s and u effects compared to the un-
substituted compound, should prove advantageous in the
event that the initial analysis indicates the unsubstituted
compound may be essentially optimal in terms of R and
u but subject to rather rapid metabolic transformation by
4-hydroxylation. The remaining listed substituents, 4-
NHCOCH3,4-NHS02CH3,4-N02,4-COCH3,4-SOzCH3,
4-CONH2,and 4-SOzNH2,are all examples of -R + u type
substituents which should prove fruitful if increased
potency is related to reduced lipophilicity or reduced li-
pophilicity combined with a +a effect. Alternatively,these
substituents may be employed in the 3 position. Cases
benefiting from a -R - u effect should have been detected
through the 4NH2and 4-OH substituents covered earlier.
The foregoing examples amply demonstrate that a
manual approach to Hansch analysis can be very useful
in efficiently identifying compounds which have a high
probability of enhanced potency in situations involving
substituent variation on a phenyl ring. The examples
considered embrace many different compound series,
Selection of Drugs for Screening Journal of Medicinal Chemistry, 1977, Vol. 20, No. 4 469

biological activities, and various types of parameter de-
pendencies. For space reasons other examples where the
method could be usefully applied have not been included.
In addition to these, some other series were found in the
literature where the activity spread between the members
of the initial compound group was too small to permit a
meaningful analysis.
A key feature is that, based on the results from only four
or five readily available analogues, the correct synthetic
direction for increased potency can often be determined.
At this stage the parameter dependency can usually be
narrowed to a small range of possibilities and further
substituents can be chosen which should increase potency
no matter what the precise activity-parameter relationship
is. This marks an important difference between the
strategy outlined in the present approach and that of the
standard Hansch method. In the latter method the object
is to first determine, utilizing a computer based analysis
of the results on eight to twelve compounds, a precise
activity-parameter relationship in the form of an equation.
This equation is then used to select new analogues which
should have improved potency. On the other hand, the
manual method does not attempt to precisely identify the
activity-parameter relationship but seeks to use a more
rapidly obtained approximate determination of this re-
lationship as a stepping stone to the identification of more
potent analogues.
In terms of numbers of compounds prepared the pos-
ition reached after preparation of the second compound
group in the manual method is roughly equivalent to that
reached after a multiple regression analysis on the first
group of compounds made in the standard Hansch ana-
lysis.24 Thus, to the extent that the present manual
method can successfully narrow the possible operative
parameter dependencies at the end of the first stage, it may
represent a more advantageous strategy if the primary goal
is to find a readily accessible compound in the maximum
potency area in the shortest possible time rather than to
determine the exact activity-parameter relationship. Also,
computers and statistical procedures are not required thus
offering greater simplicity of use for most medicinal
chemists.
References and Notes
(1) Presented in part a t the 167th National Meeting of the
American Chemical Society, Los Angeles, Calif., April 1-5,
1974,and the Fifth International Symposium on Medicinal
Chemistry, Paris, July 19-22, 1976.
(2) J. G. Topliss, J. Med. Chem., 15,1006 (1972).
(3) C. E.Granito, G. T. Becker, S. Roberts, W. J. Wiswesser,
and K. J. Windlinz, J . Chem. Doc., 11 (2),106 (1971).
(4) P. J. Goodford, Adv. Phurmucol. Chemother., 11,51(1973).
(5) C. Hansch and T. Fujita, J. Am. Chem. SOC.,86,1616(1964).
(6) C. Hansch, "Drug Design", Vol. I, E. J. Ariens, Ed., Academic
Press, New York, N.Y., 1971,p 271.
(7) P. N. Craig, J. Med. Chem., 14,680 (1971).
(8) C. Hansch, Cancer Chemother. Rep., 56,433 (1972).
(9) C. Hansch, S.H. Unger, and A. B. Forsythe, J. Med. Chem.,
16, 1217 (1973).
(10) F. Darvas, J . Med. Chem., 17,799 (1974).
(11) R.Wootton, R.Crdield, G. C. Shappey, and P. J. Goodford,
J . Med. Chem., 18,607 (1975).
(12) R. T. Buckler, S. Hayao, 0. J. Lorenzetti, L. F. Sancilio, H.
E. Hartzler, and W. G. Strycker, J. Med. Chem., 13,725
(1970).
(13) R.W. Fuller, J. Mills, and M. M. Marsh, J . Med. Chem.,
14,322 (1971).
(14) N. Kakeya, N. Yata, A. Kamada, and M. Aoki, Chem.
Pharm. Bull., 17,2558 (1969).
(15) A. Cammarata, R. C. Allen, J. K. Seydel, and E. Wempe,
J. Pharm. Sci., 59, 1496 (1970).
(16) C. Silipo and C. Hansch, Farmaco, Ed. Sci., 30,35 (1974).
(17) C. Hansch and E. W. Deutsch, Biochim. Biophys. Acta, 112,
381 (1966).
(18) C. Silipo and C. Hansch, Mol. Pharmacol., 10,954(1974).
(19) B. Blank, N.W. Di Tullio, L. Deviney, J. T. Roberts, and
H. L. Saunders, J. Med. Chem., 18,952 (1975).
(20) C. Hansch, K. H. Kim, and R. H. Sarma, J. Am. Chem. SOC.,
95,6447 (1973).
(21) M. Yoshimoto, K. N. von Kaulla, and C. Hansch, J. Med.
Chem., 18,950 (1975).
(22) G. van den Berg, T. Bultsma, R. F. Rekker, and W. T. Nauta,
Eur. J. Med. Chem., 10, 242 (1975).
(23) B. J. Broughton, P. Chaplen, P. Knowles, E. Lunt, S. M.
Marshall, D. L. Pain, and K. R. H. Wooldridge, J . Med.
Chem., 18, 1117 (1975).
(24) The option exists at this point to transpose into the standard
Hansch approach, although the compound mix may not be
optimal.

A Statistical-Heuristic Method for Automated Selection of Drugs for Screening

Louis Hodes,' George F. Hazard, Ruth I. Geran, and Sidney Richman
Division of Cancer Treatment, National Cancer Institute, Silver Spring, Maryland 20910. Received June 30, 1976
A statistical-heuristic method for selecting drugs for animal screening is developed with molecular structure features
as predictors of biological activity. The method is intended to work on large amounts of data over varied structures.
A trial of this method on a small data set allows some comparison with more sophisticated pattern recognition methods.
Problems connected with interdependence among structure predictors are critical in this method and schemes to
eliminate redundancy are reviewed. Alternate sets of structure predictors are considered. The discussion here outlines
directions to be taken in the near future.

A major activity of the Developmental Therapeutics
Program (DTP) in the Division of Cancer Treatment
(DCT), National Cancer Institute (NCI), is the develop-
ment of new drugs useful in the treatment of human
cancer. As one means of identifying leads to such drugs,
PTP, which subsumed the Drug Research and Develop-
knent Program (DR&DP), operates an antitumor screening
program that involves the testing of compounds in a
variety of animal tumor models. Because of the limited
capacity for screening, currently roughly 15 000 synthetic
compounds per year, and the almost limitless possibilities
for obtaining compounds, many approaches to selecting
acquisitions or assigning priorities of testing are being
explored.
Some of these approaches involve the use of biological
test data from previous acquisitions and chemical structure
data to create a system for predicting the biological activity
of a new compound by examining its chemical structure.'S2
Chemical structural parts are obvious choices for prediction
parameters because of their clear pertinence and easy
availability. At NCI automated files currently contain
chemical structure on more than 280 OOO compounds and